'sub' starting a word fragment

What this neuron does

This neuron fires most strongly on the piece 'sub' when it starts a longer word split across tokens, as in subcutaneous, subordinated or subunit. Turning it up made the model's text drift into unrelated topics, so it doesn't seem to steer the subject in any clear way.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

1. #1 peak

   called lean gene is not only linked to the risk of diabetes type 2 but also that of heart disease.⏎This gene IRS1 does lower fat, but only subcutaneous fat – fat under the skin, and not visceral fat, the fat

2. #2 peak

   organic chemistry laboratory for cooling reactions to sub-zero temperatures. Instead of using ice water to cool to 0 C, you

3. #3 peak

   in which the expectations are found are more⏎difficult. They will require a sophistication in manipulating⏎integrals, changing variables in integrals and using the Erf or⏎CumNorm functions. This would be a good task to use in this way

4. #4 peak

   only obstacle between the West and the fabulous contracts offered by Russia. All other matters should be subordinated to the trade—it’s a highly compelling argument.⏎The third thread paradoxically

5. #5 peak

   the plane after Kitaigorodsky . The upper arrangement has p2 plane group with twofold rotation, and the lower arrangement has pmm plane group with symmetry planes. In a complementary arrangement, the protrusion of one molecule complements

6. #6 peak

   of the muon (μ−⏎) and muon neutrino (νμ⏎) The third generation consist of the tau (τ−⏎) and the tau neutrino (ντ⏎). Each type of neutrino is associated with

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

1. #1 peak

   learn how to obtain the Java Development Kit (JDK)⏎and write your first Java program.⏎Next, it's on to the Java fundamentals, including data types,⏎operators, control statements, classes, objects, and methods.⏎You'll then progress

2. #2 peak

   -like?<|endoftext|>& Tornado Alley⏎Jan Mikusinski developed an operational calculus which is relevant for solving differential equations. His operational calculus is based upon an algebra of the convolution of functions with respect to the Fourier transform. From the convolution product he

3. #3 peak

   , which begins with the study of functions and their derivatives and integrals⏎- The fundamentals of advanced algebra, which is based on axiomatic systems involving operations of addition and multiplication in general settings⏎- Additional subjects such as geometry, probability,

4. #4 peak

   and develop an understanding of the general principles and characteristics of programming and programming languages. This includes:⏎- algorithms and problem solving with an emphasis on fundamental data structures⏎- object-oriented programming⏎- event-driven programming.⏎At least some portion of a study

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.